The I-ASD cohort was selected from the previously reported New Jersey Language and Autism Genetics Study. From three separate families, iPSCs were generated from three male probands with ASD and their sex-matched unaffected siblings (Sib) served as controls (see Materials and methods).

For comparison, we selected a genetically defined ASD cohort with 16p11.2 deletion CNV. The CNV deletes 28 genes, increases autism risk by ~20-fold, and is one of the most common ASD-associated mutations (Weiss et al., 2008; Niarchou et al., 2019; Qureshi et al., 2014). This cohort consists of three individuals, two males (16pDel M-1, 16pDel M-2) and one female (16pDel F). For sex-matched controls, we selected three iPSC lines from the NIH Regenerative Medicine Program (NIH).

We analyzed these 12 individuals (six I-ASD-Sib and six 16pDel-NIH) extensively for developmental phenotypes utilizing two to five clones per individual (except for NIH) and two to five independent neural inductions to ensure extensive rigor and reproducibility (see Materials and methods and Supplementary file 1).

WGS was conducted on I-ASD dataset and demonstrated that no individuals harbored the 16p11.2 deletion (Supplementary file 2). Moreover, no I-ASD individual had rare non-synonymous protein-coding variants in the 28 genes deleted in 16p11.2. Finally, there are no rare protein-coding non-synonymous genetic variants shared between all three I-ASD individuals. These data are consistent with the two datasets being genetically distinct with no common or shared genetic drivers.

During the mid-fetal developmental window, there are a few critical cellular processes that occur and are essential for the normal function and architecture of the brain. These processes include proliferation, migration, and differentiation. Within differentiation, neurite outgrowth is a critical cytoskeleton-dependent process necessary for normal neural connectivity in the brain. Elaboration of neurites is an important mechanism to assess for proper development in neuronal cells and multiple studies of NDDs have reported impairment in neurite outgrowth (Gleeson and Walsh, 2000; Uppal and Hof, 2013; Wegiel et al., 2010; Kathuria et al., 2018; Doers et al., 2014; Marchetto et al., 2010; Krey et al., 2013; Deshpande et al., 2017; Prem et al., 2020). Thus, we first assessed neurite outgrowth in I-ASD NPCs at 48 hr (Figure 1A and B). Given the genetic heterogeneity of the I-ASD cohort, we expected to find personalized phenotypes. Yet, surprisingly, we found that all three I-ASD NPCs had a significantly lower percentage of neurites than their respective Sibs (Figure 1B, 50–60% reduction, p<0.001 for all families). To understand how I-ASD NPCs compared to genetically unrelated neurotypical individuals, we also compared I-ASD NPCs to Sibs from other families (e.g. I-ASD-1 to Sib-2). On average, all I-ASD NPCs had reduced neurites compared to all Sib NPCs.

Given the common neurite impairment in all I-ASD NPCs, we then extended our studies to another critical cytoskeleton-dependent mid-fetal developmental process, NPC migration, which occurs concurrently with neurite outgrowth and is similarly regulated during development (Prem et al., 2020). Moreover, alterations in migration have been reported in multiple NDDs as well (Wegiel et al., 2010; Wiegreffe et al., 2015; Hori and Hoshino, 2017; Prem et al., 2020; Brennand et al., 2011b; Brennand and Gage, 2011a, Brennand et al., 2011c, Brennand et al., 2015; Pan et al., 2019). For these studies, we plated neurospheres and at 48 hr measured the area covered by migrating NPCs (Figure 1C). Remarkably, we found that all three I-ASD patients exhibited decreased migration (~30–50%) compared to their Sib (p<0.001; Figure 1D) as well as unrelated Sibs. Moreover, differences in migration area were not related to the initial neurosphere size (see Materials and methods and Figure 1—figure supplement 1). Thus, despite genetic heterogeneity both neurite outgrowth and cell migration are similarly impaired in all three I-ASD NPCs.

This remarkable convergence prompted us to examine the 16pDel ASD cohort for these phenotypes. Fascinatingly, 16pDel NPCs have a lower percentage of neurites (7.4%) than both the Sib (11.3%, p<0.001) and NIH controls (12.7%, p<0.001) (Figure 1E). This indicates that defects in neurite outgrowth are a common feature across two genetically distinct subtypes of autism. Interestingly, 16pDel NPCs do have a higher percent of neurites than I-ASD (7.4% vs 5.1%, p<0.01) indicating potential subgroup differences. There was no difference in neurite extension between Sib NPCs (11.3%) and NIH NPCs (12.7%) (p=0.1).

Lastly, we assessed migration in 16pDel NPCs. Once again, like I-ASD NPCs, all three 16pDel NPCs have impaired migration when compared to Sib and NIH NPCs (Figure 1F). There were no differences in migration between 16pDel and I-ASD or Sib NPCs and NIH NPCs.

In summary, we have found that two genetically distinct ASD cohorts have surprisingly common dysregulation in two early neurodevelopmental processes, neurite outgrowth and cell migration. Given the marked heterogeneity of ASD, these results are striking and potentially suggest that common mechanisms regulate these neurodevelopmental defects.

Few iPSC studies have explored the roles of EFs in regulating human NPC development. We postulated that stimulating NPCs with EFs could reveal defects not apparent in control conditions and help identify underlying signaling defects. Thus, we stimulated our NPCs with nerve growth factor (NGF), PACAP, and 5-HT which regulate neurite outgrowth or cell migration in rodent studies (Maisonpierre et al., 1990; Hanswijk et al., 2020; Dicicco-Bloom et al., 1998; Lu and DiCicco-Bloom, 1997). Furthermore, in prior studies in our lab, we have noted impaired response to EFs in rodent models of NDDs (Rossman et al., 2014).

We first tested EFs on the I-ASD cohort. Under PACAP, NGF, and 5-HT stimulation, Sib NPCs from all families had significant increases (45–75%, p<0.001) in neurites (Figure 2A and B, and Figure 2—figure supplement 1). In contrast, all three I-ASD NPCs failed to respond to the three EFs (Figure 2A and B, and Figure 2—figure supplement 1). The lack of response was not due to differential sensitivities or inappropriate doses of EF, as dose-response analyses indicate Sibs respond to PACAP and NGF at multiple doses while ASD had no response at any physiological dose (Figure 2—figure supplement 2).

Next, we extended EF studies to the 16pDel cohort. Surprisingly, unlike I-ASD NPCs, all three 16pDel NPCs respond to EFs with increases in neurite outgrowth. As expected, like Sib controls, NIH NPCs also exhibit increases in neurite outgrowth by EF stimulation (Figure 2C and D, Figure 2—figure supplement 3).

We also applied EFs, specifically PACAP, a well-known regulator of migration in rodent models (Falluel-Morel et al., 2005), to our migration assay. Like our neurite results, increased cell migration in Sibs was observed but there was no effect of PACAP on migration in any of the three I-ASD NPCs (Figure 2E). Likewise, all three 16pDel NPCs and NIH NPCs respond to PACAP with increased migration (Figure 2F).

In summary, while our two ASD cohorts have common impairments in neurite outgrowth and migration, EF experiments uncovered subtype-specific response in I-ASD vs 16pDel. Thus, subtype-specific defects can be present alongside common neurobiological phenotypes.

Signaling pathways are central to the regulation of neurodevelopmental processes and signaling dysregulation has been implicated in NDD pathogenesis (Lipton and Sahin, 2014; Takei and Nawa, 2014; Kelley et al., 2008; Pucilowska et al., 2012; Kwan et al., 2016; Wang et al., 2017; Waite and Eickholt, 2011). Given the neurodevelopmental phenotypes present in all our ASD NPCs and the impaired EF responses in I-ASD, we postulated that dysregulated signaling pathways could be an underlying mechanism. Thus, we conducted unbiased proteomic and p-proteomic studies comparing Sib and I-ASD, Sib and 16pDel, and I-ASD and 16pDel.

We first analyzed the proteome and found surprisingly few changes in the I-ASD and 16pDel datasets relative to Sib. Initial data included over 9700 proteins, however, to adjust for multiple comparisons we selected proteins which met significance threshold of log p>5 for analyses. With these parameters, I-ASD proteome had nine changes while 16pDel had 48 changes compared to Sib. There were only three proteins (TAPT-1, GGA-1, S100A11) in common between I-ASD and 16pDel proteomes (Figure 3A).

In striking contrast, many more changes were observed in the p-proteome analysis: I-ASD NPCs had 105 differentially phosphorylated proteins compared to Sib with 149 unique phosphorylation sites. Further, 16pDel NPCs had 419 differentially phosphorylated proteins with 916 unique phosphorylation sites. Surprisingly, given the absence of genetic overlap between the cohorts, there is considerable overlap between I-ASD and 16pDel p-proteome results with 67 p-proteins (67/105) and 107 phosphorylation sites (107/149) shared between I-ASD and 16pDel (Figure 3B). The large differences in the p-proteome along with minimal changes in the proteome suggest that phosphorylation and signaling could be dysregulated in both forms of ASD. Moreover, the p-proteome overlap between I-ASD and 16pDel NPCs suggests that different genetic mutations could converge on common signaling pathways.

To further assess p-proteome data, bioinformatic analyses were performed. To begin to understand the biological processes that could be dysregulated, G-profiler analysis was conducted (Raudvere et al., 2019; Reimand et al., 2019). For both I-ASD and 16pDel p-proteome, alterations in cytoskeleton, cell cycle, and developmental processes were identified, correlating well with our defined cellular phenotypes. To identify dysregulated signaling pathways, QIAGEN Ingenuity Pathway Analysis (IPA) was utilized (Krämer et al., 2014; Schubert et al., 2015). IPA of I-ASD p-proteome determined the top three enriched modules are tight junction signaling, mTOR signaling, and synaptogenesis signaling (Figure 3C, ). For the 16pDel p-proteome, the top three pathways are EIF2 signaling, regulation of EIF4 and p70S6K signaling, and mTOR signaling (Figure 3D, ). Comparison analysis indicated the three most overlapping pathways between I-ASD and 16pDel as mTOR signaling, telomerase signaling, and HIPPO signaling (Figure 3E).

In terms of canonical signaling pathways, mTOR has the highest enrichment in our datasets. The mTOR pathway includes a cascade of activation of molecules including phosphorylation of AKT and activation of p70S6K and ultimately phosphorylation of S6. Interestingly, IPA expression network analysis of p-proteome reveals strongly altered p-S6 as a central node of convergence in both I-ASD and 16pDel networks along with other mTOR members (Figure 3—figure supplements 1 and 2). Thus, despite the two datasets being genetically distinct, both I-ASD and 16pDel share many p-proteome changes with mTOR signaling and convergence onto RPS6 being common between them.

In striking contrast, IPA of total proteome data even with threshold log p > 3 did not find any mTOR enrichment in either I-ASD or 16pDel NPCs (Figure 3—figure supplement 3). For I-ASD, we also utilized available WGS data to run IPA. Much like the total proteome results, I-ASD WGS did not show any mTOR pathway enrichment with canonical pathway analyses or with network analysis (Figure 3—figure supplement 4). Thus, the strong converging mTOR signaling pathway abnormalities are only uncovered with analysis of the p-proteome.

To further assess the mTOR pathway, western blot studies were conducted to define p-S6 levels in 48 hr NPC cultures. Remarkably, p-S6 dysregulation was found in all six ASD NPCs from both cohorts (Figure 4). Further, less significant changes were also noted in p-AKT levels in all ASD NPCs (Figure 4—figure supplement 1). Importantly, total S6 and AKT protein levels were not different between any of the NPCs. While p-proteome analysis revealed p-S6 as a common dysregulated node among all ASD individuals (Figure 3—figure supplements 1 and 2), our western studies revealed two distinct groups of p-S6 dysregulation. In one group, p-S6 levels were reduced relative to controls whereas in the other p-S6 levels were elevated, suggesting two distinct ‘signaling subtypes’ as described below.

In the I-ASD cohort, Family-1 (I-ASD-1) had a 75% reduction in p-S6 compared to Sib-1 (Figure 4A and B) while I-ASD-3 from Family-3 had a 45% reduction in p-S6 compared to Sib-3 (Figure 4C and D, p < 0.001). Surprisingly, in contrast, I-ASD-2 from Family-2 had elevated levels of p-S6 (Figure 4E and F, 260% increase, p < 0.001) when compared to Sib-2. The abnormal p-S6 levels of each ASD NPC were also significant when compared to the mean of the three Sibs (Figure 4—figure supplement 2). Likewise, p-AKT data followed the same pattern but with a smaller effect (Figure 4—figure supplement 1). In the 16pDel group, all three 16pDel patient NPCs had higher p-S6 levels compared to both Sib and NIH NPCs (M1: ~170%, p < 0.0001, M2: ~130%, p < 0.001, F: ~80%, p < 0.01 increase, Figure 4G and H). There were no statistical differences in p-S6 levels between Sib and NIH NPCs (p = 0.99). All 16pDel NPCs also displayed higher p-AKT levels compared to NIH but not Sib NPCs (Figure 4—figure supplement 1G and H).

Thus, western studies validate the p-proteome results in our six ASD patients and confirm changes in both p-S6 and to some extent p-AKT, which can be divided into low mTOR and high mTOR groups. Ultimately, the most striking finding is the consistent mTOR signaling differences across two genetically distinct groups of autism, a finding not previously described.

To determine whether the mTOR differences described above drive the developmental abnormalities in our cohorts, we conducted gain and loss of function studies on all I-ASD and both male 16pDel NPCs. As there were no small molecule activator and inhibitor pairs for p-S6 when experiments were being conducted, we instead employed an AKT activator (SC-79) and AKT inhibitor (MK-2206), which have been used extensively in other systems to modulate mTOR pathway (Jo et al., 2012; Takizawa et al., 2012; Hirai et al., 2010). Notably, in the mTOR pathway AKT phosphorylation leads to mTOR pathway activation and ultimately S6 phosphorylation. Given p-AKT was not a central node in our p-proteome analyses and the p-AKT western data readout was less robust, our experiments largely focused on effects of small molecule activator and inhibitor on p-S6. However, we did conduct a brief proof of principle experiment which found that Sc-79 and MK-2206 did increase and decrease (respectively) AKT levels in human NPCS (data not shown).

We first examined our low mTOR cohort comprised of I-ASD-1 and I-ASD-3. Focusing first on Family-1, addition of the SC-79 activator (2 µg/mL) to I-ASD-1 NPCs increased p-S6 levels to that of Sib (Figure 5A and B). Importantly, SC-79 at this dose had no effect on p-S6 levels in Sib, nor on the levels of total S6 in either Sib or I-ASD. We next tested whether rescuing p-S6 with SC-79 affected neurite outgrowth or migration. Indeed, SC-79 increased the percentage of neurites (Figure 5C) and migration in I-ASD NPCs (Figure 5D) without changing these parameters in Sib. Thus, we could rescue the neurite and migration defects in I-ASD-1 by mTOR modulation and increasing p-S6 levels.

If mTOR pathway deficiency is truly contributing to ASD defects in neurite outgrowth and migration, then inhibitor MK-2206 should diminish neurite outgrowth and impair migration in Sib-1 NPCs. MK-2206 (30 nM) reduced p-S6 levels in Sib down to that of I-ASD-1 (Figure 5E and F). MK-2206 exposure also reduced the percentage of neurites (Figure 5G) and migration (Figure 5H) of Sib-1 NPCs to that of I-ASD-1 NPCs thereby reproducing both the ASD neurite and migration defects.

Lastly, we tested whether modulating p-S6 in I-ASD-3, the other ‘low mTOR’ individual, also led to similar results. Like in I-ASD-1, SC-79 increased p-S6 levels to that of Sib-3 in I-ASD-3 (Figure 5I). Likewise, SC-79 also led to an increase in neurite outgrowth in I-ASD-3 that paralleled Sib-3 NPCs (Figure 5J). Likewise, inhibiting p-S6 with MK-2206 in Sib-3 (Figure 5K) led to a reduction in neurite outgrowth that mimics the phenotype seen in I-ASD-3 NPCs (Figure 5L). Thus, in both ‘low-mTOR’ ASD patients, we find that increasing mTOR pathway activity (via SC-79) can rescue ASD defects while reducing mTOR pathway activity (via MK-2206) in the Sibs reproduces autism NPC phenotypes.

Given that increasing p-S6 in our low mTOR cohort led to rescue of ASD-associated phenotypes, we next sought to determine whether decreasing p-S6 in high mTOR ASD NPCs (ASD-2 and 16pDel) could similarly rescue ASD NPC phenotypes. Starting with I-ASD-2, addition of MK-2206 inhibitor reduced the levels of p-S6 to that of Sib without changing total S6 levels (Figure 6A and B). This increased both neurite outgrowth and migration paralleling levels seen in Sib, thereby rescuing the phenotypes (Figure 6C and D). Significantly, Sib-2 had no significant response to MK-2206 at this dose.

We next determined if elevating mTOR signaling using SC-79 could reproduce the I-ASD-2 defects in the Sib. SC-79 increased p-S6 levels in Sib-2 NPCs to that of I-ASD-2 (Figure 6E and F), which remarkably led to neurite and migration impairments that mimic those seen in I-ASD-2 (Figure 6G and H).

Finally, we conducted gain and loss of function studies in the two male 16pDel NPC and NIH NPCs. Like I-ASD-2, the 16pDel NPCs have reduced neurites and migration in conjunction with elevated p-S6. Reducing p-S6 levels with inhibitor MK-2206 (Figure 6I) rescued the 16pDel phenotypes as shown by increased neurite outgrowth (Figure 6J). Conversely, increasing p-S6 with activator SC-79 (Figure 6K) in NIH NPCs reproduced the neurite defects noted in the 16pDel NPCs (Figure 6L). In summary, we find that both over- and under-activation of mTOR signaling leads to common reduction in neurites and migration. Manipulation of mTOR could rescue or reproduce ASD phenotypes in both low and high mTOR NPCs, indicating that tight regulation of mTOR pathway is essential for normal neural development.

The above results indicate that mTOR signaling dysregulation is responsible for the shared neurodevelopmental phenotypes. However, one phenotype that diverges between I-ASD NPCs and 16pDel NPCs is responsiveness to EFs. Unlike Sib, NIH, and 16pDel NPCs, the I-ASD NPCs do not respond to EFs (PACAP, NGF, and 5-HT) which stimulate neurite outgrowth and migration in all other NPC subgroups. Given that PACAP, NGF, and 5-HT act through different receptor systems we postulated that dysregulated signaling could potentially contribute to lack of I-ASD response to EFs. Thus, to further investigate if this failure to respond to EFs is due to mTOR signaling defects, we treated the I-ASD NPCs with ‘subthreshold’ amounts of mTOR agonist and inhibitor which did not maximally increase p-S6 or elicit neurite outgrowth on its own. To do so, we performed parallel dose-response studies and selected drug doses that did not elicit neurite outgrowth on their own but did elicit a non-maximal p-S6 increase (data not shown). For Family-1, agonist SC-79 (0.1 µg/mL) did not stimulate neurite outgrowth in I-ASD-1 NPCs (Figure 7A). However, when combined with PACAP, NGF, and 5-HT, the I-ASD-1 NPCs now responded to all three EFs (Figure 7A). This rescue effect was specific to I-ASD-1 NPCs as addition of subthreshold SC-79 to Sib NPCs did not heighten EF response (Figure 7B). Thus, by using subthreshold doses of mTOR activator, we could facilitate EFs signaling in I-ASD-1 NPCs.

Conversely, to test whether diminishing p-S6 in Sib abolishes EF responses, we took a similar ‘subthreshold’ approach. Addition of ‘subthreshold’ MK-2206 (1 nM) to Sib-1 NPCs abolished their typical EF responses (Figure 7C).

Finally, we also examined high mTOR Family-2: using a ‘subthreshold’ dose of inhibitor (1 nM MK-2206), I-ASD-2 NPCs were now able to respond to EFs (Figure 7D). These ‘subthreshold’ studies further support altered mTOR signaling in NPCs and that manipulation of mTOR signaling can alter ASD and control NPC response to EFs during neurodevelopment.

While most ASD iPSC studies have investigated terminal neuronal differentiation phenotypes, our study has focused on NPCs and may be the first to find common NPC neurite and migration impairments in two different ASD subtypes. Previous iPSC-based studies of NDDs, such as Fragile X syndrome, 22q13 deletion syndrome, 16p11.2 deletion and duplication syndromes, and macrocephalic I-ASD patients have noted alterations in dendrites, axons, and neurites in post-mitotic neurons (Bhattacharyya and Zhao, 2016; Castrén et al., 2005; Doers et al., 2014; Shcheglovitov et al., 2013; Kathuria et al., 2018; Deshpande et al., 2017; Griesi-Oliveira et al., 2015; Marchetto et al., 2017). Migration defects have not been commonly noted in ASD but have been reported in studies of NDDs including schizophrenia, Rett syndrome, Down syndrome, Timothy syndrome, and CNTNAP2 mutation (Brennand et al., 2015; Flaherty et al., 2017; Krey et al., 2013; Huo et al., 2018; Zhang et al., 2016; Birey et al., 2022). Thus, our studies provide further evidence to altered developmental neuronal phenotypes in ASD.

Multiple studies indicate that NPCs are a biologically relevant cell type for ASD. Genetic studies, including extensive WES and GWAS analysis, have reproducibly identified human mid-fetal cerebral/cortical (8–24 weeks) development as a window for when and where most ASD risk genes are expressed (Satterstrom et al., 2020; Willsey et al., 2013; Parikshak et al., 2013). During this window, NPCs are formed and undergo migration and early neurite extension. Recently altered NPC development has also been identified as a consistent neurodevelopmental phenotype for numerous ASD risk genes (Garcia-Forn et al., 2020). Given that human iPSC-derived NPCs are similar transcriptionally and functionally to cortical NPCs, our studies provide experimental evidence of NPC developmental dysregulation in ASD (Hofrichter et al., 2017; Yan et al., 2013a). Furthermore, by comparing two subtypes of ASD, we show, as suggested by genetics studies, that this early developmental dysregulation may be a common feature among multiple forms of ASD.

Our p-proteome, westerns, and small molecule analyses identified altered mTOR signaling as a point of convergence for the shared ASD neurite and migration phenotypes. To our knowledge, we are the first to utilize unbiased p-proteomics in ASD NPCs and were able to uncover marked differences in the p-proteome for both ASD subtypes when compared to control. Interestingly, despite non-overlapping genetics, significant overlap was observed (67/105) at the p-proteome level between I-ASD and 16pDel. Further, about a third of these p-proteins are common with 102 high confidence ASD risk genes or relevant gene families (Satterstrom et al., 2020), suggesting that we are identifying similar pathways as described by others. In contrast, very few proteomic changes were noted in I-ASD (9/9700) and 16pDel (48/9700) NPCs when compared to unaffected individuals. This could suggest either the proteome is relatively unaltered in ASD NPCs, or the changes are quite varied between samples so no consistent significant results were observed.

The convergent signaling results are even more remarkable when we consider the WGS results and total proteome results. There is no rare variant overlap between the two ASD datasets, or among the three I-ASD individuals. While some I-ASD individuals have rare variants in genes that could affect mTOR signaling, these were often also shared by the Sib (e.g. TSC-1 rare variants). Given that the non-overlapping variants could potentially converge onto the mTOR pathway, we subsequently ran IPA on WGS data which did not uncover convergence onto the mTOR pathway or any other classical signaling pathway such as ERK, WNT, or CREB, but did find common alterations in immune regulatory pathways (Figure 3—figure supplement 4). Likewise, pathway analysis of published transcriptome data from 16pDel ASD patient iPSCs or lymphoblastic tissue also reveals no mTOR enrichment or dysregulation in signaling (Blumenthal et al., 2014). These results are also mirrored in the total proteome analyses which do not suggest mTOR enrichment (Figure 3—figure supplement 3). These results suggest that even without genetic mutations or protein changes in the mTOR pathway, multiple rare and common variant genes could interact to lead to dysregulation of mTOR pathway signaling. Moreover, the presence of high and low mTOR I-ASD NPCs could conceivably be dependent upon the specific type of genetic mutations present and their effect on mTOR signaling. Consistent with this possibility, all three 16pDel individuals who share the same CNV deletion have NPCs with high mTOR signaling. Ultimately, genetic analyses alone did not reveal convergence onto the mTOR pathway or alteration in the signaling milieu in our cohort. We were only able to detect our remarkably convergent mTOR deficits and alterations in signaling by utilizing p-proteomics and studying cellular signaling through western blotting, further indicating that ASD could be a disorder of intracellular signaling.

Consistent with the idea that mTOR signaling dysfunction could be a common mechanism for autism, monogenic syndromic disorders, which can exhibit ASD such as tuberous sclerosis (TSC), PTEN-associated ASD, and neurofibromatosis-1, have mutations that directly affect mTOR pathway components (Enriquez-Barreto and Morales, 2016; Yeung et al., 2017; Huber et al., 2015; Dasgupta et al., 2005; Winden et al., 2019; Sharma et al., 2010; Rangasamy et al., 2016; Costales and Kolevzon, 2015). Furthermore, some studies of syndromic diseases where the driving mutation is not in the mTOR pathway, such as Fragile-X, Angelman, Rett syndrome, and Phelan-McDermid syndrome (22q13 deletion), also show mTOR alterations in human postmortem samples, embryonic stem cell models, and hiPSC-based models (Winden et al., 2018). In 16pDel, because the MAPK3 gene is deleted, both rodent and human studies have uncovered alterations in ERK1 signaling (Pucilowska et al., 2018; Pucilowska et al., 2015; Deshpande et al., 2017). Importantly for 16pDel, altered mTOR has not been reported previously. On the other hand, there are no iPSC studies of I-ASD that show mTOR abnormalities. However, recent studies in peripheral white blood cells identified mTOR signaling proteins and p-proteins as being predictors of not only ASD diagnosis but also disease severity (Onore et al., 2017; Rosina et al., 2019). Yet, it is unclear whether WBC mTOR alterations are sufficient to suggest that mTOR is contributing to neurodevelopmental phenotypes in these I-ASD individuals. Regardless, our study adds to the growing literature that mTOR signaling dysfunction is a shared pathogenic mechanism among different types of ASDs and raises the possibility that mTOR dysregulation could be observed commonly across multiple additional ASD subtypes. Thus, studying mTOR in multiple ASD subtypes could be important in deciphering the molecular etiology of ASD.

Of course, while mTOR seems to play an important role in multiple subtypes of ASD, numerous other signaling pathways have also been implicated in the pathogenesis of ASD. As noted above, in 16p11.2 deletion, ERK/MAPK3 is part of the 11.2 segment and implicated in driving some of the neurodevelopmental phenotypes associated with the deletion and duplication syndromes. The WNT pathway, which includes CHD8 (an autism-related chromatin modifier) and beta-catenin, has long been implicated in ASD pathogenesis. The WNT pathway, much like mTOR, is thought to regulate neuronal proliferation and has been associated with macrocephalic phenotypes seen in ASD. For example, a study by Marchetto et al. found that NPCs derived from individuals with macrocephaly and I-ASD displayed excessive proliferation as well as altered WNT signaling. In our p-proteome analyses, the ERK/MAPK and PKA pathway have also been shown to be enriched. Ultimately, altered cellular signaling seems to play an important role in the pathogenesis of ASD and other NDDs. Importantly, our study indicates that ASD NPCs not only have enrichment of dysregulated mTOR but that mTOR dysregulation is driving autism-associated phenotypes.

The importance of mTOR signaling was further highlighted by our small molecule gain and loss of function analyses and EF studies. By correcting the p-S6 levels with mTOR agonists and antagonists, we were able to rescue the developmental defects in both I-ASD and 16pDel NPCs. Furthermore, by changing Sib control p-S6 to ASD levels, the neurodevelopmental defects could be reproduced demonstrating that the mTOR signaling differences are mechanistically linked to the shared neurodevelopmental defects for both I-ASD and 16pDel NPCs. In addition, as one of the first studies to challenge ASD NPCs with EFs, we were able to uncover the importance of mTOR in the regulation of the signaling milieu and uncover differences between I-ASD and 16pDel in their ability to respond to EFs. Specifically, by utilizing subthreshold doses of SC-79 and MK-2206, we were able to facilitate responses to EFs in all I-ASD NPCs. These results suggest that mTOR homeostasis and ambient levels may be critical for facilitating several cellular processes including the ability for other signaling pathways to function effectively. As 16pDel NPCs had no EF response impairments, different ASD subtypes or different genetic backgrounds may have distinct thresholds for maintaining signaling homeostasis. Our studies point strongly to the importance of the signaling milieu and mTOR regulation in the control of neurodevelopment and ASD pathogenesis.

Interestingly, in I-ASD NPCs, mTOR seems to play a pivotal role in EF responsiveness. Both ‘high’ and ‘low’ I-ASD NPCs fail to respond to EFs while altering mTOR milieu with subthreshold doses of MK-2206 or SC-79 allows the I-ASD NPCs to respond to EFs. Yet, remarkably, 16pDel NPCs, which share elevated mTOR signaling with I-ASD-2 NPCs, have no deficits in EF responsiveness. This suggests that mTOR dysregulation alone is not sufficient to cause EF deficits in all NPCs. While there are several commonalities between 16pDel NPCs and I-ASD NPCs, our proteomic analyses show that there are significant differences as well. Specifically, the I-ASD p-proteome has 38 differently phosphorylated proteins that are not shared with the 16pDel and the 16pDel p-proteome has 351 differently phosphorylated proteins not shared with the I-ASD NPCs. Analysis of these ‘differently’ expressed p-proteins show that I-ASD NPCs seem to have dysregulation in tight junction signaling as well as in proteins such as spectrin and ankyrin which are essential to appropriate stabilization of membrane proteins (such as signaling receptors) to the underlying cytoskeleton. Thus, it is possible that interaction of mTOR defects with tight junction/anchoring molecules lead to the lack of responses to numerous EFs (all of which work on different receptor systems). We have not conducted specific studies into the difference between I-ASD and 16pDel when it comes to EF responsiveness, though the p-proteome results provide the above tantalizing clues which could be focused on in future research. Regardless, the most interesting point is that despite many commonalities, I-ASD and 16pDel NPCs do not share the EF unresponsiveness phenotype, though for I-ASD NPCs, EF responsiveness can be altered by targeting mTOR signaling.

In the developing brain, EFs regulate basic developmental processes like neurite outgrowth and migration and alterations in EFs have been implicated in the pathogenesis of NDDs (Otto et al., 2001; Hashimoto et al., 2001; Cameron et al., 2007; Lu and DiCicco-Bloom, 1997; Yan et al., 2013b: DiCicco-Bloom, 1996; Nicot and DiCicco-Bloom, 2001; Korsching et al., 1986; Rajakumar et al., 2004; Lazar et al., 2008; Bonnin et al., 2007; Temple and Qian, 1995; Speranza et al., 2015; Boukhris et al., 2016; Veenstra-VanderWeele et al., 2012; Yang et al., 2014). In our study, we selected three EFs, PACAP, NGF, and 5-HT, known to be expressed during the ‘critical window’ of mid-fetal development, with well-established roles in regulation of early neurodevelopmental phenotypes to stimulate I-ASD and 16pDel NPCs in both our neurite outgrowth and neurosphere assays to assess for ASD-specific response differences. Strikingly, NPCs derived from all three idiopathic patients are unresponsive to all three EFs tested, while all the unaffected patient NPCs had increased neurite outgrowth or migration in response to EFs. Such differential response to EFs has been noted in previous work in our lab on the Engrailed-2 KO model of developmental disorders (Rossman et al., 2014). Though studies are limited, further exploration may show that differential responses to regulatory molecules may be a characteristic of some types of developmental disorders. Indeed, meta-analyses have found that about 30% of children with ASD have elevations in blood 5-HT. This increased 5-HT could be due to dysregulated 5-HT production, enhanced uptake, impaired clearance, or a compensatory response to decreased sensitivity to 5-HT in the ASD brain. At the same time, our studies found that unlike I-ASD NPCs, 16pDel NPCs have a typical response to all three EFs with increases in neurite outgrowth and migration. Response to EFs could also have implications for the treatment of individuals with ASD, for example, SSRI are commonly used to treat comorbid depression/anxiety in ASD, but if these individuals have neural cells that are less responsive to 5-HT, this treatment could be less effective, though further research is needed to understand whether NPC responses to EFs/drugs correlate with drug responses in the individual. In sum, while different subtypes of ASD can have common underlying neurobiology, there are subtypes or perhaps even patient-specific differences in the defects seen in ASD. By conducting experiments with EFs we were able to find two distinct subgroups of ASD which would not have been apparent if we had conducted our experiments in control conditions alone, showing the utility of studying patient-derived cells under different conditions.

The heterogeneity of ASD, the subjective nature of clinical categorization, and the current inability to stratify and subtype the disorder have been postulated as reasons for the decades of unsuccessful clinical trials in ASD. Effectively subtyping ASD would help us uncover the multiple pathways through which ASD can manifest in different individuals, allow us to take a more ‘personalized’ approach to etiological studies, and could help us understand which therapeutics would be best suited to certain ASD subpopulations. As our study is one of the first to compare two disparate forms of ASD, it has given us insight into both convergent and personalized phenotypes in ASD populations.

Despite common neurodevelopmental phenotypes, both a ‘low’ and ‘high’ mTOR signaling subtype is present in our ASD NPCs. This phenomenon of ‘too much’ or ‘too little’ of a pathway or gene leading to similar outcomes has been seen before in MECP-2 gain and loss of function as well as in Timothy syndrome (Birey et al., 2022; Peters et al., 2013; Ramocki et al., 2010). Our gain and loss of function studies showed that mTOR dysfunction drives impaired neurodevelopment. Thus, characterizing ASD by mTOR levels could help us identify which groups may be candidates for mTOR activators vs inhibitors in future drug trials. Another subtype-specific difference was that I-ASD individuals failed to respond to EFs while 16pDel ASD individuals had typical increases in neurites and migration under EF stimulation. This difference in EF response did not correlate with the mTOR defects. Specifically, both I-ASD-2 and 16pDel had ‘high’ mTOR signaling but only I-ASD-2 failed to respond to EFs, indicating that our EF studies help us glean more information than molecular studies alone. While there are no medications that target the core symptoms of ASD, we do use drugs that modulate monoamines such as serotonin (5-HT) and dopamine to manage certain behavioral symptoms in ASD. Furthermore, comorbid conditions such as depression are common in ASD and notoriously difficult to treat with our usual medications (Williams et al., 2010). Thus, understanding whether certain subtypes of ASD are responsive to EFs like 5-HT or even medications by using iPSC-derived cells may help us better tailor treatments for ASD and its comorbid conditions.

Our study is one of the first to show the remarkable convergence of neurodevelopmental and molecular phenotypes in two distinct autism subtypes using rigorous methods such as utilization of multiple iPSC clones and neural inductions. The extensive rigor of our analyses that included a total of 29 distinct iPSC clones and 61 distinct neural inductions lends to relatively small sample sizes (though similar to other iPSC studies) with only six autism probands, three of each subtype. Thus, it is difficult to know whether our observed phenotypes are generalizable to all individuals with 16pDel, I-ASD, or other ASDs. However, it is important to note that our sample had significant clinical and genotypic heterogeneity yet, we could still uncover common phenotypes. Expanding our study design to other subtypes of ASD or other NDDs would help establish if common underlying mechanisms could be identified (Gandal et al., 2018).

While our studies reveal important alterations in early neurodevelopment in ASD, our monolayer NPC studies do not recapitulate the 3D nature or complex interactions occurring in the developing brain. Moreover, as hiPSC-derived NPCS are most similar to fetal cortical radial glial cells, the phenotypes observed in our dishes may not directly parallel the structural or functional defects in the ASD brain. Neuropathological studies of ASD have indicated brain defects suggestive of dysregulated development. However, studies have not been conducted to determine whether, for example, a defect in migration in ASD iPSC-derived NPCs is correlated with altered cortical lamination in the patient from whom the cells were derived. To understand this, MRI and postmortem analyses need to be conducted to correlate iPSC results with neuropathological alterations. Interestingly, an iPSC study of ASD patients with MRI-defined macrocephaly found that affected NPCs proliferated faster than control NPCs (Marchetto et al., 2017; Connacher et al., 2022b), suggesting correlation between structural MRI and NPC phenotypes.

Our data potentially suggests that mTOR could serve as a target to employ precision medicine techniques in ASD (Sato, 2016). However, given that NPCs that are fetal in nature, it is unclear whether targeting mTOR in a child or adult would treat disease phenotypes. Importantly, however, we know that signaling pathways such as mTOR continue to play roles in the postnatal and adult brain (Pagani et al., 2021). For example, mTOR is critical for the regulation of processes such as the moment-to-moment modification of dendritic spines which are essential to learning, memory, and integration of information in the brain (Ganesan et al., 2019; Nicolini et al., 2015; Tordjman et al., 2015; Henry et al., 2017). Thus, it is reasonable to postulate that mTOR defects, which in the prenatal period leads to altered neurites or migration, could in the adult brain cause altered dendritic functioning and plasticity which could consequently lead to behavioral symptoms of ASD. As such, targeting mTOR in the adult brain could alleviate ASD symptoms. Preclinical mouse models of Rett syndrome support this concept, as treatment of adult animals with mTOR-activating IGF-1 reduced anxiety levels and increased exploratory behaviors, suggesting that targeting mTOR can have an effect outside of the fetal developmental window (Castro et al., 2014). On the other hand, clinical trials of IGF-1 in children with Rett syndrome and rapamycin in children with TSC have not thus far had promising effects on behavior (Khwaja et al., 2014, Sahin M, 2018b, Sahin M, 2018a). While early neurodevelopmental processes may no longer be occurring in the postnatal brain, studying NPCs can provide insight into the initial processes that contribute to ASD pathogenesis.

Sib and I-ASD iPSCs were generated by the Lu & Millonig labs while the 16pdel iPSCs were generated by SAMPLED. For I-ASD and Sib, T lymphocytes were isolated from cryopreserved samples and these cells were expanded and then reprogrammed with a non-integrating Sendai virus containing the four Yamanaka factors: SOX2, OCT3/4, KLF, and a temperature-sensitive C-MYC (Seki et al., 2011). The 16pdel iPSCs were also made using the same Sendai virus method, however, both lymphocytes and fibroblasts were used as the initial somatic tissue. iPSCs were characterized using immunocytochemistry (ICC) and QRTPCR for the following markers: NANOG, OCT4, TRA-1–60, SSEA4, CD24, and E-Cadherin. iPSCs were assayed for chromosomal abnormalities via G-band karyotype assay or via Array Comparative Genomic Hybridization (Company: Cell Line Genetics); DNA fingerprinting was also conducted to confirm genetic identity to original tissue samples. See Connacher et al., 2022a, Materials and methods, and supplementary information section for details and data on these iPSC characteristics. iPSC lines were maintained in mTeSR1 media (STEMCELL Technologies, 85850) on six-well plates (Corning, COR-3506) coated with hESC qualified Matrigel (Corning, 354277). mTeSR media was treated with Primocin (antimicrobial) at 1:500 dilution. Media was changed daily. Once cells reached 70–90% confluency, they were passaged by incubating cells with 0.5 mM filter-sterilized EDTA diluted in 1× PBS for 10–20 min. When iPSCs detached from the plate, they were lifted and centrifuged at 150×g for 5 min. The cell pellet was then gently resuspended in mTeSR and plated at 15–30% confluence in mTeSR media with 5 µM rock inhibitor (Y-compound), Y27632 (STEMCELL Technologies, 72302) for 24 hr.

To generate NPCs, iPSCs were induced using a modified version of the Thermo Fisher GIBCO Neural Induction Protocol (Williams et al., 2018). In brief, iPSCs at 70–90% confluency were dissociated using 1× Accutase (Thermo Fisher Scientific A111050) and were centrifuged at 150×g for 5 min. The resulting pellet was then resuspended into Neural Induction Media (NIM) (Thermo Fisher Scientific, A1647801) with 5 µM of Y-compound. The following densities of iPSCs were plated into Matrigel-coated 12-well plates containing 1 mL of NIM (with 5 μM Y-compound): 80, 120, 200, 300 K. Two wells of each density were made. One set of wells at each density were induced for 7 days while the other was induced for 8 days. Once induction day 7 or 8 was reached, cells were lifted with 1× Accutase and centrifuged at 300×g and then resuspended in Neural Expansion Media (NEM) (Thermo Fisher Scientific, A1647801). 1–1.5 million cells were then replated onto Matrigel-coated six-well plate with 2 mL NEM with 5 µM Y-compound. At this point cells are P0. For our studies, NPCs were used only from P3 to P8. At P3-P5 lines were tested with Lonza mycoplasma detection kit to ensure that media/cells were contamination free. All media utilized was supplemented with primocin (antimicrobial) at 1:500 dilution.

To assess NPC identity, ICC was conducted to verify that cells expressed the following NPC markers (Sox2, Pax6, and Nestin). NPCs were discarded if marker immunostaining revealed Nestin or Sox2 <85% or Pax6 <60% (Connacher et al., 2022a). NPCs were also stained with OCT3/4 to ensure they no longer expressed this iPSC marker. As shown in Connacher et al., 2022b, Quantiplex analysis was also performed on NPC mRNA to assess expression of several NPC markers.

NPCs were also differentiated into neurons, astrocytes, and oligodendrocytes as per Thermo Fisher Protocol: https://www.thermofisher.com/us/en/home/references/protocols/neurobiology/neurobiology-protocols/differentiating-neural-stem-cells-into-neurons-and-glial-cells.html. Differentiated cell identity was assessed by ICC for the following markers: MAP2/Tau (Neuron), O4/Gal-c (oligodendrocytes), and GFAP for astrocytes (Connacher et al., 2022a).

WGS data for the samples in the I-ASD dataset was extracted from a previous study (Zhou et al., 2023). The details of sequencing and variant identification are described previously (Zhou et al., 2023). The sequencing data and the variant calls are available in the National Institute of Mental Health Data Archive (NDA) under projects C1932 and C2933.

The variants were annotated with ANNOVAR (ver. 2017-07-17, human reference build hg19) (Wang et al., 2010) to obtain the mutation effect, population allele frequency (AF), etc. After annotation, variants were selected based on the following criteria: (1) have a ‘PASS’ value in the FILTER field; (2) have an AF <5% in database gnomAD_genome_ALL; (3) exonic variants with the functional categories of ‘non-synonymous’, ‘frameshift_deletion’, ‘frameshift_insertion’, ‘nonframeshift_deletion’, ‘nonframeshift_ insertion’, ‘stopgain’, and ‘stoploss’; (4) not in dispensable genes as defined by MacArthur et al., 2012, and Rausell et al., 2020.

IPA was also applied to the WGS data. For detailed methods on IPA please see the Phospho-proteome section of the Materials and methods. For IPA WGS, briefly, gene variations only found in I-ASD, with functional impact, and with a read quality of ‘pass’, were isolated. Then, for each I-ASD family, variants present only in the affected individual were extracted. The Gene ID for this variant and Gnomad frequency were uploaded to IPA. Core analysis was completed the same as for the p-proteome data.

Cells were fixed in 4% PFA for 15 min and then permeabilized with 0.3% Triton X-100 in PBS for 10 min. Then, cells were blocked with 10% Normal Goat Serum for 1 hr and incubated in the appropriate primary antibodies from mouse or rabbit: Sox2 (1:1000, Abcam), Oct4 (1:250, Santa Cruz, Sc-5279), Nestin (1:500, R&D Systems, MAB1259), Pax6 (1:300, Covance, PRB-278P), β-III tubulin (TuJ1, 1:200 Covance, MMS-435P), Tau (1:500, Santa Cruz, Sc-5587), O4 (R&D Systems, MAB1326 1:500), GalC (Santa Cruz 1:250 sc-518055), and Glial fibrillary acidic protein (GFAP, 1:1000, Dako, G9269). Then, cells were washed three times for 5 min with 1× PBS and incubated 1 hr in appropriate secondary antibody at 1:1000. The secondary antibodies used were red (594 Alexa Fluor, Thermo Fisher Scientific, Invitrogen, A-11032, A-11037) or green (488 Alexa Fluor, Thermo Fisher Scientific, Invitrogen, A-11001, A-11070) goat-anti-mouse antibodies or goat-anti-rabbit antibodies. Cells were visualized on fluorescent microscope.

The protocols for all experimental media conditions, neurite outgrowth, and neurosphere assays are described in detail in our prior JOVE paper (Williams et al., 2018).

All experiments were conducted in 30% neural expansion media (30 NEM) which was made by diluting NEM by 70% in a 1:1 ratio of DMEM/F12 and NB media. Primocin antibiotic (100 µg/mL, InvivoGen ant-pm-1) was also added to the media.

50,000 NPCs were plated onto a 35 mm dish coated with 0.1 mg/mL PDL and 5 µg/mL Fibronectin (Sigma-Aldrich F1141) containing 1 mL of 30 NEM without or with drugs/EFs. Vehicle, EF, or small molecule inhibitor/activator were added to media prior to cell plating. The EFs and concentrations used include: NGF (Peprotech, AF-450-01) at 3, 10, 30, and 100 ng/mL dissolved into 30 NEM; PACAP (American Peptide/BACHEM, H-8430) at 1, 3, 10, 30, and 100 nM dissolved into 30 NEM; and Serotonin (Sigma-Aldrich, H9523) at 30, 100, 300 µg/mL dissolved into 30 NEM. The small molecule activator and inhibitor and concentrations were: SC79 (AKT activator, S7863) in DMSO diluted into 30 NEM at concentration of 0.1, 0.3, 1, 3, 6 µg/mL and MK-2206 (AKT inhibitor, S1078) in DMSO diluted into 30 NEM at concentrations of 0.3, 1, 3, 10, 30, and 100 nM. Vehicles were matched to whichever carrier was used to dissolve the drug or EF (e.g. for small molecule studies, vehicle was DMSO dissolved into 30 NEM and volumes equivalent to the drug were added to each dish). Drugs and EFs were not replenished, and media was not changed in the 48 hr period of incubation and as such the initial EFs/molecules added at the start of the experiment remained for the full 48 hr period.

For each condition (control or EF or small molecule), two to three dishes were set up in each experiment. NPCs were incubated at 37°C for 48 hr and then fixed in 4% PFA. In each dish, the proportion of cells bearing neurites was counted blind. Neurites were defined as processes that extend from the cell body that are equal to or greater than 2 cell body diameters in length. For cells with more than one process, the longest process was assessed. Cells were counted directly on a phase contrast microscope at 32×. Two to four randomly chosen 1 cm rows were counted per dish. At least 150 cells were counted per dish and the proportion of cells with neurites was calculated in each dish.

Neurospheres were formed by dissociating confluent NPCs and plating 1 million cells into uncoated 35 mm dishes with 100 NEM. No small molecules or EFs were added during this formation step. NPCs were then incubated at 37°C for 24–96 hr (varied from line to line) to allow aggregation into neurospheres. Sphere size was assessed daily using a live-ruler on a phase contrast microscope. When a majority of spheres reached an approximate diameter of 100 μm (±20 μm), the migration assay was performed.

To coat plates, a Matrigel aliquot was dissolved into 6 mL of ice-cold 30 NEM. Appropriate vehicles and EFs were added to the 30 NEM/Matrigel mixture as desired. For migration the following EFs and small molecule inhibitors were used as noted above in the neurite section (PACAP, SC-79, MK-2206). Then, 1 mL of the 30 NEM/Matrix ± EF mixture was added into the six-well plate. Plates were then incubated for 30 min at 37°C. While plates were incubating, neurospheres were collected from the 35 mm dishes and placed into a 15 mL conical tube. The tube was centrifuged at 100×g and the pelleted spheres were gently resuspended using a P1000 in 1–3 mL of pre-warmed 30 NEM (with appropriate concentration of EF or small molecule added). 200 μL of the sphere suspension was then placed into the 30 NEM Matrigel plates. Spheres were allowed to migrate for 48 hr and were fixed for 30 min in 4% PFA, washed, and stored in PBS + 0.05% sodium azide.

Neurospheres were imaged at 10× on a phase contrast microscope. At least 15 neurospheres across all dishes were imaged per condition. Spheres that were not in contact with another exhibited a contiguous migrating cell carpet and intact inner cell mass were imaged. Average migration was measured using ImageJ. To measure migration, the outer contour of a neurosphere was traced using the freehand line tool and the enclosed area was measured. Then, the inner cell mass area was measured. Migration was quantified by subtracting the inner cell mass area from the total neurosphere area. At least 15 neurospheres were analyzed per condition. Exclusion criteria are detailed in JOVE methods paper (Williams et al., 2018).

Given that neurospheres are not the same size when plated, it was important to ensure that initial sphere size (ISS) did not affect migration. As such, proof of principal tests were conducted to assess if ISS influenced migration. Migration and ISS data were obtained from each ‘diagnosis’ group (Sib, NIH, I-ASD, 16pASD) over two experiments with 15 neurospheres each. Linear modeling was conducted in R studio utilizing the basic equation lm(Migration~ISS±diagnosis) where ‘lm’ allows R to conduct linear modeling, ~ allows for R to calculate the relationship between migration and ISS utilizing any variables and data specified. Results of these analyses and others are found in the supplement, but ultimately, these studies showed us that the ISS did not affect migration, whereas diagnosis did (Figure 1—figure supplement 1). Thus, for all remaining neurosphere studies, ISS was not factored into analyses.

To collect protein, confluent NPCs were dissociated, pelleted, and plated onto PDL + Fibronectin-coated plates with 1 mL 30 NEM. 1 million NPCs were plated in 35 mm dishes and incubated for 48 hr at 37°C. At 48 hr, cells were treated with vehicle or drug (SC-79 or MK-2206) for 20 or 30 min. Then, dishes were washed with ice-cold PBS two times. Cells were then lysed by adding 30 μL of ice-cold lysis buffer/dish and scraping with cell scraper. Lysate was then sonicated 2× for 1 min on ice and then samples were centrifuged at 4°C for 10 min. The supernatant was removed and saved while the pellet was discarded. Protein was aliquoted and stored at –80°C.

For small molecule drugs utilized (SC-79 and MK-2206), studies in other cellular systems had shown appropriate increase and reduction in p-S6, respectively, with drug (Hao et al., 2016; Zhu et al., 2019; Chen et al., 2017; Simioni et al., 2013; Somnay et al., 2013; Wilson et al., 2014; Luan et al., 2018). However, at the time of our studies, there had been no published papers showing the effects of these drugs on human NPCs. As such, our initial studies with these molecules involved confirming the effect of SC-79 and MK-2206 on p-AKT and p-S6 levels in our NPCs. We initially focused our studies on two I-ASD pairs, Family-1 in which I-ASD NPCs had reduced p-S6 levels and Family-2 in which I-ASD NPCs had increased p-S6 levels. We trialed multiple concentrations (SC-79: 0.1, 0.3, 1, 2, 3, 10 µg/mL, MK-2206, 0.3, 1 3, 10, 30, 100 nM) of each small molecule drug as well as multiple time points (5 min, 10 min, 20 min, 30 min, 1 hr, 24 hr, and 48 hr) to arrive at the doses and 20 and 30 min stimulation paradigm applied in this manuscript (data not shown). Given thorough testing in these two families we were able to clearly establish that SC-79 increases both p-AKT and p-S6 and MK-2206 decreases both p-AKT and p-S6 in Sib-1, I-ASD-1, Sib-2, and I-ASD-2. As such, in the other NPCs (Family-3, 16pDel and NIH), we did not conduct rigorous tests with these small molecules.

Aliquoted samples were thawed and 20 μg of protein was mixed with appropriate volumes of 4× NuPAGE LDS Sample buffer and 10× NuPAGE Sample Reducing Agent. All immunoblotting reagents were obtained from Thermo Fisher Scientific. Samples were boiled, cooled, and loaded into wells on a 12% SDS-PAGE poly-acrylamide gel. The samples were run at 100 V for 1.5 hr in an electrophoresis apparatus with 1× NuPAGE Mops SDS Running Buffer. Protein was then transferred to a PVDF membrane using a wet transfer apparatus with 1× NuPAGE Transfer buffer. The membranes were washed with Tris-buffered saline (TBS) with Tween 20 (0.1%) (TBS-T) and blocked with 5% powdered milk in 0.1% TBS-T. Then, membranes were probed for proteins with the following antibodies used at a 1:2000 concentration and incubated overnight at 4°C: S6 (2317, Cell Signaling), p-S6 (2211, Cell Signaling) as previously reported (Genestine et al., 2015; Yan et al., 2013b, Mairet-Coello et al., 2009; Mairet-Coello et al., 2012; Tury et al., 2011). GADPH was used as a loading control and was probed with antibody (Meridian Life Sciences, H86045M) at a dilution of 1:10,000 for 1 hr. After incubation at 1°C, membrane was washed and appropriate HRP-conjugated 2°C antibody was applied at a 1:1000 concentration for all antibodies except GAPDH, for which 2°C antibody was at 1:5000. Then, ECL was applied to membranes for 1 min and membrane was applied to medical grade X-ray film. For most experiments, two gels were loaded with identical protein samples and then each gel was transferred to a separate membrane. For experiments requiring fewer samples, a single gel was run and then cut in half prior to transfer. Multiple wells were loaded, often with protein from different clones and different neural inductions. While each of these gels had identical samples loaded, one gel was utilized to probe for all phospho-proteins and the other for total protein to avoid stripping and to decrease background signal. Membranes were often probed with multiple antibodies. Each primary was incubated on its own for 2 hr at room temperature or overnight at 4°C, followed by the appropriate secondary and ECL steps. Film exposure time varied by antibody and often multiple films were exposed to chemiluminescent bands and then the films that were not over or undersaturated were selected for scanning into JPEG. After the above process was completed for one antibody of interest, the film was washed for 1 hr with TBS and then reprobed with another primary antibody. Often, the first antibody utilized was the one that had the weakest signal. For our experiments the order of antibodies was as follows: p-AKT, p-S6, and then GAPDH. Scanned films were then quantified on ImageJ. Band intensities for phospho- and total protein were normalized to signals for GAPDH. The GAPDH-normalized phospho-antibody intensity was divided by the GAPDH-normalized total antibody to get a relative protein intensity.

Data acquired from the idiopathic autism individuals were compared to data acquired from unaffected Sibs. Data was compiled and organized in several ways. Comparisons were made within Family (Sib-1 vs ASD-1), across families (Sib-1 vs ASD-2), or as an average of all Sib (three patients, all clones) compared to the average data from all idiopathic ASD (three patients, all clones). For every neurite experiment, two to three dishes were set up per condition. For every migration condition, 20–30 spheres were evaluated per condition. Dishes were kept as separate N-values. For example, when compiling data for Sib-1 in control condition, all neurite dishes from every clone were averaged together to get a mean neurite value for this individual but SEM was calculated utilizing each dish as a separate ‘N’ value. For spheres, each neurosphere was considered a separate statistical point. Migration values from each sphere derived from multiple clones were averaged together to acquire an average migration value for each patient. Spheres were also analyzed by clone. For the 16p11.2, data was compared to composite averages of the Sibs (dishes from all three Sib patients and all clones) or averages of the NIMH controls (dishes from both patients, all neural inductions). Assuming statistical normality, in comparisons that include only two groups, unpaired t-test was used to test differences in affected vs unaffected samples. Simple unpaired or paired t-tests were run in Microsoft Excel. If normality assumption was not satisfied, the non-parametric Wilcoxon test was used in GraphPad Prism 8. The presence of multiple ASD subtypes and the use of EFs lead to instances where multiple group comparison is necessary. For such cases, in normal data, analysis of variance or ANOVA (one or two way) was used to detect statistical significance. To reduce type I error, p-values for the ANOVAs were calculated using Tukey correction in GraphPad Prism 8. If normality assumption did not hold, non-parametric Kruskal-Wallis test was applied for multiple comparisons.